I modified the dcgan in examples for my application. I tried to run the script on gpu server. After training 3 mini-batches the training stops and logout from gpu server with the error “Write failed: Broken pipe”. I didn’t encounter this kind of error before, can you please tell me the problem. You can find the training code here https://codeshare.io/2EoKzp
Typical SSH kick, not PyTorch related. Try using
ssh -o ServerAliveInternal=60 username@server
These lines are already in my ~/.ssh/config
I get the error only when I try to run the program.
I recently encountered the same problem only when I run some pytorch code. Have you already solved this?